ext2: Move direct-io to use iomap

This patch converts ext2 direct-io path to iomap interface.
- This also takes care of DIO_SKIP_HOLES part in which we return -ENOTBLK
  from ext2_iomap_begin(), in case if the write is done on a hole.
- This fallbacks to buffered-io in case of DIO_SKIP_HOLES or in case of
  a partial write or if any error is detected in ext2_iomap_end().
  We try to return -ENOTBLK in such cases.
- For any unaligned or extending DIO writes, we pass
  IOMAP_DIO_FORCE_WAIT flag to ensure synchronous writes.
- For extending writes we set IOMAP_F_DIRTY in ext2_iomap_begin because
  otherwise with dsync writes on devices that support FUA, generic_write_sync
  won't be called and we might miss inode metadata updates.
- Since ext2 already now uses _nolock vartiant of sync write. Hence
  there is no inode lock problem with iomap in this patch.
- ext2_iomap_ops are now being shared by DIO, DAX & fiemap path

Tested-by: Disha Goel <disgoel@linux.ibm.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Message-Id: <610b672a52f2a7ff6dc550fd14d0f995806232a5.1682069716.git.ritesh.list@gmail.com>

1 files changed, 115 insertions, 0 deletions

diff --git a/fs/ext2/file.c b/fs/ext2/file.c
index 749163787139..98add36c1a59 100644
--- a/fs/ext2/file.c
+++ b/fs/ext2/file.c
@@ -162,12 +162,124 @@ int ext2_fsync(struct file *file, loff_t start, loff_t end, int datasync)
 	return ret;
 }
 
+static ssize_t ext2_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
+{
+	struct file *file = iocb->ki_filp;
+	struct inode *inode = file->f_mapping->host;
+	ssize_t ret;
+
+	inode_lock_shared(inode);
+	ret = iomap_dio_rw(iocb, to, &ext2_iomap_ops, NULL, 0, NULL, 0);
+	inode_unlock_shared(inode);
+
+	return ret;
+}
+
+static int ext2_dio_write_end_io(struct kiocb *iocb, ssize_t size,
+				 int error, unsigned int flags)
+{
+	loff_t pos = iocb->ki_pos;
+	struct inode *inode = file_inode(iocb->ki_filp);
+
+	if (error)
+		goto out;
+
+	/*
+	 * If we are extending the file, we have to update i_size here before
+	 * page cache gets invalidated in iomap_dio_rw(). This prevents racing
+	 * buffered reads from zeroing out too much from page cache pages.
+	 * Note that all extending writes always happens synchronously with
+	 * inode lock held by ext2_dio_write_iter(). So it is safe to update
+	 * inode size here for extending file writes.
+	 */
+	pos += size;
+	if (pos > i_size_read(inode)) {
+		i_size_write(inode, pos);
+		mark_inode_dirty(inode);
+	}
+out:
+	return error;
+}
+
+static const struct iomap_dio_ops ext2_dio_write_ops = {
+	.end_io = ext2_dio_write_end_io,
+};
+
+static ssize_t ext2_dio_write_iter(struct kiocb *iocb, struct iov_iter *from)
+{
+	struct file *file = iocb->ki_filp;
+	struct inode *inode = file->f_mapping->host;
+	ssize_t ret;
+	unsigned int flags = 0;
+	unsigned long blocksize = inode->i_sb->s_blocksize;
+	loff_t offset = iocb->ki_pos;
+	loff_t count = iov_iter_count(from);
+
+	inode_lock(inode);
+	ret = generic_write_checks(iocb, from);
+	if (ret <= 0)
+		goto out_unlock;
+
+	ret = kiocb_modified(iocb);
+	if (ret)
+		goto out_unlock;
+
+	/* use IOMAP_DIO_FORCE_WAIT for unaligned or extending writes */
+	if (iocb->ki_pos + iov_iter_count(from) > i_size_read(inode) ||
+	   (!IS_ALIGNED(iocb->ki_pos | iov_iter_alignment(from), blocksize)))
+		flags |= IOMAP_DIO_FORCE_WAIT;
+
+	ret = iomap_dio_rw(iocb, from, &ext2_iomap_ops, &ext2_dio_write_ops,
+			   flags, NULL, 0);
+
+	/* ENOTBLK is magic return value for fallback to buffered-io */
+	if (ret == -ENOTBLK)
+		ret = 0;
+
+	if (ret < 0 && ret != -EIOCBQUEUED)
+		ext2_write_failed(inode->i_mapping, offset + count);
+
+	/* handle case for partial write and for fallback to buffered write */
+	if (ret >= 0 && iov_iter_count(from)) {
+		loff_t pos, endbyte;
+		ssize_t status;
+		int ret2;
+
+		iocb->ki_flags &= ~IOCB_DIRECT;
+		pos = iocb->ki_pos;
+		status = generic_perform_write(iocb, from);
+		if (unlikely(status < 0)) {
+			ret = status;
+			goto out_unlock;
+		}
+
+		iocb->ki_pos += status;
+		ret += status;
+		endbyte = pos + status - 1;
+		ret2 = filemap_write_and_wait_range(inode->i_mapping, pos,
+						    endbyte);
+		if (!ret2)
+			invalidate_mapping_pages(inode->i_mapping,
+						 pos >> PAGE_SHIFT,
+						 endbyte >> PAGE_SHIFT);
+		if (ret > 0)
+			generic_write_sync(iocb, ret);
+	}
+
+out_unlock:
+	inode_unlock(inode);
+	return ret;
+}
+
 static ssize_t ext2_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
 {
 #ifdef CONFIG_FS_DAX
 	if (IS_DAX(iocb->ki_filp->f_mapping->host))
 		return ext2_dax_read_iter(iocb, to);
 #endif
+	if (iocb->ki_flags & IOCB_DIRECT)
+		return ext2_dio_read_iter(iocb, to);
+
 	return generic_file_read_iter(iocb, to);
 }
 
@@ -177,6 +289,9 @@ static ssize_t ext2_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
 	if (IS_DAX(iocb->ki_filp->f_mapping->host))
 		return ext2_dax_write_iter(iocb, from);
 #endif
+	if (iocb->ki_flags & IOCB_DIRECT)
+		return ext2_dio_write_iter(iocb, from);
+
 	return generic_file_write_iter(iocb, from);
 }